Inappropriate Exclusion of Organic Solvents

[momeara]

The standardiser inappropriately removes organic compounds that are sometimes used as organic solvents. For example, the important metabolite glutamate is removed:

from rdkit import Chem
from standardiser import standardise
smiles = 'NC(CCC(=O)O)C(=O)O' # glutamate
mol = Chem.MolFromSmiles(smiles, sanitize=False)
clean_mol = standardise.apply(mol) # throws no_non_salt exception

One approach to resolving this issue would be allow organic solvents if they are the only organic fragment in the molecule. Here is code that does this:

    organic_salt_frags = []

    non_salt_frags = []

    mol = break_bonds.apply(mol)

    for n, frag in enumerate(Chem.GetMolFrags(mol, asMols=True), 1):

        logger.debug("Starting fragment {n} '{smi}'...".format(n=n, smi=Chem.MolToSmiles(frag)))

        logger.debug("1) Check for non-organic elements...")

        if unsalt.is_nonorganic(frag): continue

        logger.debug("2) Attempting to neutralise (first pass)...")

        frag = neutralise.apply(frag)

        logger.debug("3) Applying rules...")

        frag = rules.apply(frag, output_rules_applied=output_rules_applied, verbose=verbose)

        logger.debug("4) Attempting to neutralise (second pass)...")

        frag = neutralise.apply(frag)

        logger.debug("5) Checking if frag is a salt/solvate...")

        if unsalt.is_salt(frag):

            organic_salt_frags.append(frag)

            continue

        logger.debug("...fragment kept.")

        non_salt_frags.append(frag)

    if len(non_salt_frags) == 0:

        logger.debug("No non-salt frags. If there is a single organic salt frag take it.")

        if len(organic_salt_frags) == 1:

            parent = organic_salt_frags[0]

        else:

            raise StandardiseException("no_non_salt")

    elif len(non_salt_frags) > 1:

        raise StandardiseException("multi_component")

    else:

        parent = non_salt_frags[0]

[flatkinson]

In the original use-case, solvent etc. were not of interest so were to be eliminated from the input. I suspect that would still be the most common case, so I think keeping them should be a deliberate choice. I could make this behaviour optional. Alternatively, allowing a custom salt dictionary could be an option.

[UnixJunkie]

Somewhat related: rejected molecules get either no_non_salt or multi_component.
I am OK with the multi_component case.
However, it seems that non_non_salt is thrown even in the case where the problem with the molecule is something else (like known organic solvent, reported above).
As a curious user, who is not a chemist, I would appreciate if the software was more specific about the error related to each molecule.
Also, it would be nice to specify via command line options, that a given problematic category should not be filtered out.